Sentinel lymph node sampling has become an alternative to axillary lymph node dissection to provide prognostic and treatment information in breast cancer patients. The role of immunohistochemistry has yet to be established. A total of 241 sentinel lymph nodes (in 270 slides) from 91 patients with invasive carcinoma (73 ductal, 9 lobular, 8 mixed lobular/ductal, 1 NOS) were studied for presence of macrometastases (> 0.2 cm), identified in hematoxylin and eosin sections, and occult metastases (micrometastases [< or = 0.2 cm], clusters of cells, isolated carcinoma cells), identified only by immunohistochemistry. Intraoperative touch preparations, frozen sections, seven hematoxylin and eosin levels (L1-L7), and two AE1-3 cytokeratin immunohistochemistries (L1, L4-5) of the entire bisected or trisected sentinel lymph node were examined. Thirty-one (34%) patients had 50 positive sentinel lymph nodes. Twenty-six (33%) sentinel lymph nodes had metastatic carcinoma (11 macrometastases, 11 micrometastases, 3 clusters of cells, 1 isolated carcinoma cells) by touch preparations, frozen sections, and one hematoxylin and eosin (L1). Thirty-eight (43%) were positive by AE1-3 immunohistochemistry (L1) (11 macrometastases, 8 micrometastases, 13 clusters of cells, 6 isolated carcinoma cells), significantly more than by touch preparations, frozen sections, hematoxylin and eosin L1, or hematoxylin and eosin L2-7. Cytokeratin immunostain on L4-5 demonstrated 31 (34%) positive sentinel lymph nodes, a similar frequency to cytokeratin immunostain on L1. Size of sentinel lymph node metastasis did not correlate with size, histologic grade, or type of primary breast carcinoma. AE1-3 (L1) immunohistochemistry is highly sensitive in delineating sentinel lymph node metastasis, especially clusters of cells and isolated carcinoma cells. The prognostic significance of clusters of cells and isolated carcinoma cells and the value of AE1-3 immunohistochemistry on frozen sections need to be determined.